github-jwinterm/xmr-stak
1
0
Fork 0
mirror of https://github.com/jwinterm/xmr-stak.git synced 2026-10-17 05:41:41 +02:00
Code Issues Packages Projects Releases Wiki Activity

fix nvidia auto suggestion

Browse Source 
The lmem is still incalculably and crash the miner very often.
Increase the potential lmem usage to 16kiB to respect lmem alignments, ...
This commit is contained in:
psychocrypt
2017-11-17 10:04:16 +01:00
parent 51c7b26e7e
commit d326644d8b
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
xmrstak/backend/nvidia/nvcc_code/cuda_extra.cu
+2 -2
View File
@@ -467,9 +467,9 @@ extern "C" int cuda_get_deviceinfo(nvid_ctx* ctx)
// 200byte are meta data memory (result nonce, ...)
size_t availableMem = freeMemory - (128u * byteToMiB) - 200u;
size_t limitedMemory = std::min(availableMem, maxMemUsage);
// up to 1kibyte extra memory is used per thread for some kernel (lmem/local memory)
// up to 16kibyte extra memory is used per thread for some kernel (lmem/local memory)
// 680bytes are extra meta data memory per hash
size_t perThread = hashMemSize + 1024u + 680u;
size_t perThread = hashMemSize + 16192u + 680u;
size_t max_intensity = limitedMemory / perThread;
ctx->device_threads = max_intensity / ctx->device_blocks;
// use only odd number of threads